Class actions allege that thousands of patients were recorded by ambient AI without consent. Whatever a court decides, an organization needs to be able to show what its AI actually did.
This page is an illustration. It reconstructs one scenario from public reporting, with synthetic identifiers and a sample receipt. It was not issued by the live engine, and nothing below is a live verification. To verify a real engine-signed receipt, use the verify page.
Governance systems deployed across fintech, mortgage, automotive, manufacturing, banking, insurance, and other regulated sectors.
Illustrative engine run logIllustration
01Inputambient_scribe.session_startcaptured
02ContextJurisdiction resolved · US-CAall-party
03GovernCIPA §632.7 consent artifactMISSING
04GovernCMIA §56.101 authorizationMISSING
05DecideVerdict renderedBLOCK
06SignEd25519 · chain committedSAFE
07OutputDoCR · docr_2d7f…6f03done · 142ms
govtrace-engine · governed7/7 steps · complete
§01What Public Reporting Tells Us
The scenario class: ambient AI consent failures.
Nov 2025 · San Diego Superior Court
Saucedo v. Sharp HealthCare
Class action alleges ambient AI recorded a July 2025 patient visit without consent. EHR notes reportedly contained boilerplate language stating the patient had been "advised" and "consented." Per the complaint, no such consent occurred.
Apr 2026 · N.D. California
Sutter Health & MemorialCare
Proposed nationwide class covering patients over the prior two years. Alleges clinicians "intercepted, recorded, and processed" audio without informed consent. Same ambient documentation category as the Sharp matter.
2026 · Multi-State Exposure
13 all-party consent states
Florida F.S. §934.03 makes unconsented recording a third-degree felony, up to five years. A single ambient scribe workflow can be lawful in one state and criminal in another.
250+
Health systems reported to use the vendor's ambient AI
100K+
Patients in the alleged class
$5K
CIPA damages · per violation
5 yrs
Maximum prison term under Florida F.S. §934.03 (third-degree felony)
Sources: Medscape (Jan 16 & Apr 16, 2026); Becker's Hospital Review (Dec 18, 2025); Fisher Phillips (Dec 9, 2025); ABA Health Law (Feb 10, 2026); HealthLaw Attorney Blog (Feb 23, 2026).
§02The Evidence Gap
What a court may ask these systems to produce.
01
A signed, tamper-evident record that consent was captured before recording began.
Not a chart note. Not a checkbox. A cryptographically signed event with timestamp, clinician ID, policy version, and patient ID. Produced at runtime, independently verifiable without contacting the issuer.
02
Proof that jurisdiction-aware policy was enforced at the moment of decision.
All-party consent in CA, FL, and 11 other states. Standard HIPAA elsewhere. The same ambient workflow deployed nationwide cannot self-attest which policy governed which visit on which date.
03
An audit chain that cannot be rewritten after the fact.
EHR notes can be edited. Vendor logs are controlled by the vendor. Neither may count as independent evidence when the consent workflow itself is the defendant.
This is not a detection problem. DLP catches data leaving. GRC platforms score posture. AI governance dashboards observe behavior. They are not designed to produce a signed runtime receipt for a consent event.
§03The Counterfactual Artifact
The Duty-of-Care Record GoVTraceAI would have generated.
An illustration of the record GoVTraceAI is designed to issue, built on a reconstructed scenario matching the cited complaints. The values below are synthetic.
Scenario reconstructed from publicly reported complaints. Identifiers, hashes and the signature are synthetic. The structure follows the GoVTrace Receipt format; this exact receipt was not signed by the engine. The Duty-of-Care Record is signed as a GoVTrace Receipt (format spec).
§04Verify This Receipt
See how verification works. Then verify a real receipt.
This block is a simulation. It runs in your browser against a stored fixture, with a timer standing in for the verifier, so it proves nothing about any receipt. The real verifier checks an Ed25519 signature against GoVTraceAI's published public key: verify a real receipt here.
Illustration only. Production verification is done by the GoVTraceAI runtime, not by this page.
§05Where Existing Tools Stop
The category gap.
DLP / Data Security
Data loss prevention tools
Typically classify and block data movement. Not designed to represent a consent event or sign a runtime decision record.
GRC / Compliance
Compliance automation platforms
Typically point-in-time posture attestation, not per-event runtime evidence.
AI Governance
AI governance and monitoring tools
Typically observability dashboards and model risk scores, not signed per-decision records.
EHR Audit Logs
EHR audit logs
Record what clinicians did in the system. They do not cover what third-party AI did before data entered the chart.
GoVTraceAI
A signed record for regulated AI.
A runtime decision engine that evaluates every AI action against policy and issues a signed, verifiable Duty-of-Care Record before data leaves your systems. Built by Gobots.
§06Run it on your scenario
Want to see this against a scenario you actually worry about?
Bring your ambient scribe workflow. Your patient-messaging copilot. Your inbox AI. Your prior-auth model. 20 minutes, live, on your scenario. You'll see the DoCR your compliance team should already have.